Projects/Codes

The related paper (our model name Perception)

https://arxiv.org/abs/2110.10965

This repository contains the code of our model in CATARACTS Semantic Segmentation 2020.
Team Perception are Mobarakol Islam, Bharat Giddwani and Ren Hongliang. They used an encoder-decoder archi- tecture for segmentation. The model was adopted from their previous works [5], [6] which contains a residual encoder and a Skip-competitive Spatial and Channel Squeeze & Excitation (SC-scSE) decoder as shown in Figure 5-a. The encoder is formed by 5 residual layers as ResNet18 and the corresponding decoding block contains convolution, adaptive batch normal- ization [7], SC-scSE, and deconvolution sequentially. The SC- scSE decoder retains weak features, excites strong features and performs dynamic spatial and channel-wise feature recal- ibration which makes the network capable of better feature learning. They used batch size of 10 for training the proposed model. The model was trained with a learning rate of 0.0001, using the Adam optimizer and the momentum and weight decay set as constant to 0.99 and 10−4, respectively. The input images were flipped randomly as a part of augmentation. They followed the same data split as [4] for train and validation.

Mengya Xu, Mobarakol Islam, Lim Chwee Ming, Hongliang Ren

We develop a multi-layer transformer based model with the gradient reversal adversarial learning to generate a caption for the multi-domain surgical images that can describe the semantic relationship between instruments and surgical region of interest (ROI).

Preprint | Paper | Code

Mobarakol Islam, Navodini Wijethilake, Hongliang Ren

In this paper, we propose a radiogenomic overall survival (OS) prediction approach by incorporating gene expression data with radiomic features such as shape, geometry, and clinical information. We exploit TCGA (The Cancer Genomic Atlas) dataset and synthesize the missing MRI modalities using a fully convolutional network (FCN) in a conditional Generative Adversarial Network (cGAN).

Preprint | Paper | Code

Mobarakol Islam, Vibashan VS, Lim Chwee Ming, Hongliang Ren.

We propose an end-to-end trainable Spatio-Temporal Multi-Task Learning (ST-MTL) model with a shared encoder and spatio-temporal decoders for the real-time surgical instrument segmentation and task-oriented saliency detection.

Preprint | Paper | Code

Mobarakol Islam, Lalithkumar Seenivasan, Lim Chwee Ming, Hongliang Ren.

We develop an approach to generate the scene graph and predict surgical interactions between instruments and surgical region of interest (ROI) during robot-assisted surgery.

Preprint | Paper | Presentation| Code

Mobarakol Islam, Vibashan VS, Hongliang Ren.

We develop a novel end-to-end trainable real-time Multi-Task Learning (MTL) model with weight-shared encoder and task-aware detection and segmentation decoders. Optimization of multiple tasks at the same convergence point is vital and presents a complex problem.

Preprint | Paper | Presentation| Code

Mobarakol Islam, Yueyuan Li, and Hongliang Ren.

We propose an end-to-end trainable multitask learning (MTL) model for real-time surgical instrument segmentation and attention prediction. Our model is designed with a weight-shared encoder and two task-oriented decoders and optimized for the joint tasks.

Preprint | Paper | Code

Mobarakol Islam, Daniel Anojan Atputharuban, Ravikiran Ramesh, Hongliang Ren.

We have developed a light-weight cascaded convolutional neural network (CNN) to segment the surgical instruments from high-resolution videos obtained from a commercial robotic system.

Preprint | Paper | Code

Mobarakol Islam, VS Vibashan, V Jeya Maria Jose, Navodini Wijethilake, Uppal Utkarsh, Hongliang Ren.

We have developed a light-weight cascaded convolutional neural network (CNN) to segment the surgical instruments from high-resolution videos obtained from a commercial robotic system.

Preprint | Paper | Code

Mobarakol Islam, VS Vibashan, V Jeya Maria Jose, Hongliang Ren.

Segmentation of brain tumor from magnetic resonance imaging (MRI) is performed using a convolutional neural network (CNN) with hypercolumn technique. Also, a variety of features are extracted from the segmented tumor to predict the overall survival in terms of number of days for each patient.

Preprint | Paper | Poster

Jia Yi Lee, Mobarakol Islam, Jing Ru Woh, T S Mohamed Washeem, Lee Ying Clara Ngoh, Weng Kin Wong, Hongliang Ren.

In this paper, we propose a tracking-by-segmentation model with spatial and channel ‘Squeeze and Excitation'(scSE) for US needle detection and trajectory prediction. We adopt a light deep learning architecture (e.g., LinkNet) as our segmentation baseline network and integrate the scSE module to learn spatial information for better prediction.

Preprint | Paper | Code

Mobarakol Islam, Parita Sanghani, Angela An Qi See, Michael Lucas James, Nicolas Kon Kam King, Hongliang Ren.

We develop a deep learning approach for automated intracerebral hemorrhage (ICH) segmentation from 3D computed tomography (CT) scans. Our model, ICHNet, evolves by integrating dilated convolution neural network (CNN) with hypercolumn features where a modest number of pixels are sampled and corresponding features from multiple layers are concatenated.

Preprint | Paper | Base Code